1405.6360.pdf

Design of a Scalable Hybrid MAC Protocol forHeterogeneous M2M Networks

Yi Liu, Chau Yuen, Senior Member, IEEE, Xianghui Cao, Member, IEEE, Naveed Ul Hassan, andJiming Chen, Senior Member, IEEE

AbstractA robust and resilient medium access control (MAC)protocol is crucial for numerous machine-type devices to concur-rently access the channel in a machine-to-machine (M2M) network.Simplex (reservation- or contention-based) MAC protocols arestudied in most literatures which may not be able to provide ascalable solution for M2M networks with large number of hetero-geneous devices. In this paper, a scalable hybrid MAC protocol,which consists of a contention period and a transmission period,is designed for heterogeneous M2M networks. In this protocol,different devices with preset priorities (hierarchical contendingprobabilities) rst contend the transmission opportunities followingthe convention-based -persistent carrier sense multiple access(CSMA) mechanism. Only the successful devices will be assigneda time slot for transmission following the reservation-based time-divisionmultiple access (TDMA)mechanism. If the devices failed incontention at previous frame, to ensure the fairness among alldevices, their contending priorities will be raised by increasingtheir contending probabilities at the next frame. To balance thetradeoff between the contention and transmission period in eachframe, an optimization problem is formulated to maximize thechannel utility by nding the key design parameters: the contentionduration, initial contending probability, and the incremental indi-cator. Analytical and simulation results demonstrate the effective-ness of the proposed hybrid MAC protocol.

Index TermsHomogeneous and heterogeneous networks,hybrid medium access control (MAC), machine-to-machine(M2M) networks.

I. INTRODUCTION

I NTERNET of Things (IoT) is an integrated part of futureinternet including existing and evolving internet and net-work developments and could be conceptually dened as adynamic global network infrastructure with self-conguringcapabilities [1][3]. Machine-to-machine (M2M) represents a

future IoTwhere billions to trillions of everyday objects and thesurrounding environment are connected and managed through arange of devices, communication networks, and cloud-basedservers. M2M communication is dened as the informationexchange between machines and machines without any humaninteraction. M2M network is expected to be widely utilized inmany elds of pervasive IoT applications [4][9], includingindustrial and agricultural automations, health care, transportsystems, and electricity grids. Recently, the enormous economicbenets of the M2M communications drive intensive discussionin international standardization activities, such as local thermalequillibrium (LTE) and IEEE 802.16. There are several char-acteristics of M2M networks:1) massive number of devices in service coverage and con-

current network access attempt from these devices;2) high level of system automation in which the devices and

systems can exchange and share data;3) heterogeneous quality of service (QoS) in M2M network

that may require priority-based medium access control(MAC) protocol;

4) fairness concern for different devices to share or competethe limited resources;

5) the battery-powered devices may need energy-efcientaccess and transmission control mechanism;

6) the data packet may be dropped due to time sensitivity ofthe data readings, infrequency, and small burst transmis-sion in low deploy cost M2M network [10].

In order to handle the massive access in M2M, 3GPP LTE hasseveral work items dened on M2M communications, primarilywith respect to overload control [14], [15]. IEEE 802.16pproposals addressed enhancements for IEEE 802.16 m [16],[17] standard to support M2M applications. It is noted that themassive access management of M2M communication overwireless channels generally happen at the MAC layer. Hence,the design of a smart and efcient MAC protocol remains a keyrequirement for successful deployment of any M2M networks.As discussed by 3GPP and IEEE 802.16, the MAC protocol forM2M networks focused on contention-based random access(RA) schemes [19][21] that allow all of the devices contendthe transmission opportunities in entire frame. The contention-based RA is popular due to its simplicity, exibility, and lowoverhead. Devices can dynamically join or leave without extraoperations. However, the transmission collisions are eminentwhen huge number of M2M devices tries to communicate thebase station (BS) all at once. One of the solutions is to usereservation-based schemes such as time-division multiple access(TDMA) inM2Mnetwork. The TDMA scheme is well known asthe collision-free access scheme where the transmission time

Manuscript received October 31, 2013; accepted February 27, 2014. Date ofpublicationMarch 11, 2014; date of current versionMay 05, 2014. Thisworkwassupported in part by programs ofNSFCunderGrant 61370159, Grant U1035001,Grant U1201253,Grant 61203117, andGrant 61203036, in part by the SingaporeUniversity Technology and Design under Grant SUTD-ZJU/RES/02/2011, inpart by National Program for Special Support of Top-Notch Young Professionalsunder Grant NCET-11-0445, and in part by Lahore University of ManagementSciences (LUMS) via Faculty start-up Grant.

Y. Liu is with the GuangdongUniversity of Technology, Guangzhou 510006,China, and also with the Singapore University of Technology and Design,Singapore 138682 (e-mail: [email protected]).

C. Yuen is with the Singapore University of Technology and Design,Singapore 138682 (e-mail: [email protected]).

N. U. Hassan is with the Department of Electrical Engineering, SSE, LahoreUniversity of Management Sciences (LUMS), Lahore 54792, Pakistan (e-mail:[email protected]).

X. Cao and J. Chen are with the State Key Laboratory of Industrial ControlTechnology, Department of Control, Zhejiang University, Hangzhou 310027,China (e-mail: [email protected]; [email protected]).

Color versions of one ormore of the gures in this paper are available online athttp://ieeexplore.ieee.org.

Digital Object Identier 10.1109/JIOT.2014.2310425

IEEE INTERNET OF THINGS JOURNAL, VOL. 1, NO. 1, FEBRUARY 2014 99

2327-4662 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

is divided into slots and each device transmits only during itsown time slots [22]. The main defect of TDMA is the lowtransmission slot usage if only a small portion of devices havedata to transmit.Hence, the pure contention-based or reservation-based scheme

may not be suitable to build up a scalable, exible, and automaticcommunication structure for a dense heterogeneous M2M net-work. In [23] and [24], the researchers introduced a hybridscheme which attempt to combine the best features of both ofreservation-based and contention-based while offsetting theirweaknesses. In [23], the authors designed the hybrid MACscheme, for sensor network to adapt to the level of contentionin the network under low contention; it behaves like carrier sensemultiple access (CSMA), andunder high contention, likeTDMA.In [24], the authors proposedhow touse hybridMACprotocols tosupport video streaming over wireless networks. Such schemestried to adapt to different bandwidth conditions depending ondemand. In addition, some existing standards, such as IEEE802.15.3, IEEE 802.15.4, and IEEE 802.11ad [25][27], adoptthe hybrid MAC protocols to satisfy the advance wireless net-working throughput.However, fewof the existing researches andstandards of the hybrid MAC protocol consider the heteroge-neous applications for different users which may exist in M2Mnetworks. In addition, considering the particular characteristicsof the M2M network such as the massive access control, energyefciency, and fairness, a new hybrid MAC protocol should bedesigned and evaluated.In this paper, we rst develop a hybrid MAC protocol for

heterogeneous M2M networks, which will combine the benetof both contention-based and reservation-based protocols. In thisprotocol, the contention and reservation process of differentdevices is a frame procedure which composes of two portions:contention only period (COP) and transmission only period(TOP). The COP is based on -persistent CSMA mechanismwhich allows different devices to contend transmission slots withtheir own priorities, i.e., the contending probabilities. Onlysuccessful contending devices are allowed to transmit dataduring TOP that provides TDMA type of data communication.To ensure the fairness, if the devices failed in contention at theprevious frame, their contending priorities will be raised byincreasing the contending probabilities at the next frame. Giventhe frame duration, it is expected that the number of successfuldevices increases when the COP duration is prolonged. How-ever, the COP duration increases at the expense of shortening theTOP, which results in the decrease of transmission slots. Toachieve the optimal tradeoff between the contention andtransmission period in each frame, an optimization problem isformulated to maximize the channel utility by deciding theoptimal contending probability during COP, and the optimalnumber of devices allowed transmitting during TOP (which isrelated to the duration of COP). The analytical and simulationresults demonstrate the effectiveness of the proposed hybridMAC protocol.In summary, we make the following contributions in this

paper:1) We design a scalable hybrid MAC protocol incorporating

with -persistent CSMA mechanism and TDMA mecha-nism for heterogeneous M2M networks.

2) By dening different contending priorities, the protocolis able to allow the heterogeneous devices to obtainhierarchical performances.

3) We report an incremental contention priority method toguarantee the access fairness of devices, which failed tocompete the transmission opportunities over time.

4) We identify the key design parameters to achieve maxi-mum channel utility by considering the contending priorityas well as the trade-off between contending period andtransmission period.

The remainder of this paper is organized as follows. In Sec-tion II, we describe the systemmodel of anM2M network. Then,we design a scalable hybrid MAC protocol in Section III. Byoptimizing the duration of COP and TOP, we introduce a hybridaccess control scheme in Section IV. Performance study andevaluationaregiven inSectionV.SectionVIconcludes thepaper.

II. M2M NETWORK MODEL

In this paper, we consider a heterogeneous M2M network,which consists of oneBSand number of devices ,as shown in Fig. 1. In this network, different types of devicesare categorized into the priority classes , eachof which has number of devices and

. Without loss of generality, we assume that thehigher classes of the devices have higher requirements of thetransmission performance, such as higher channel utility andlower packet drop ratio, than that of the lower classes of devices.BS is responsible for operating MAC for different classes ofthe active devices (the devices that have packet to transmit) byassigning different contending probabilities to class

, respectively. We assume that the device withhigher priority has higher contending probability than that withlower priority, i.e., < < .1 To specically describethe contending priorities of different classes of devices,we assumethat the contending probabilities have therelationship as

where is dened as the incremental indicator.For each active device, the data packet arrival process is

modeled as a Possion arrival process with packet arrival rate .

Fig. 1. System model of M2M network.

1We can also set the priority to the packets instead of devices, e.g. some packetshave higher priority than the others.

100 IEEE INTERNET OF THINGS JOURNAL, VOL. 1, NO. 1, FEBRUARY 2014

Here, for simplicity, we assume that all devices have the samepacket arrival rate, which is known by BS. A new packet thatarrives at a device is buffered until the device successfullycontends the transmission opportunity and nishes the transmis-sion. If a new packet arrives at the device before the transmission,the buffered packet will be replaced by the new one. Hence, thereis one packet at most in the buffer of each device. Moreover, inorder tomaintain stationary property,we assume that all devices inthe network are static (i.e., we do not consider mobility effects).

III. A SCALABLE HYBRID MAC PROTOCOL DESIGN

In this section, we design a scalable hybrid MAC protocol forheterogeneous M2M network. We consider the operation of theM2M network on a frame-by-frame basis. Each frame is com-posed of four portions as depicted in Fig. 2: notication period(NP), COP, announcement period (AP), and TOP. During NP,the BS broadcasts notication message to all devices for notify-ing the beginning of the contention. The active devices will con-tend the channel during COP. The COP is based on -persistentCSMA access method [29], and is used for devices to randomlysend the transmission request to BS. After COP, the BS broad-casts the beginning of the transmission period during AP. Thedevices succeeded in contention is allowed to transmit datapacket during the remaining time of a frame, which is speciedas the TOP. The TOP provides a TDMA type of communicationfor the devices. We assume that each assigned transmission slothas the same length and there is no transmission error for eachdevice [11], [12]. The specic description of the BSs anddevicess operations in each period is given as follows.

A. Operation of BS and Devices

1) Notication Period: At the start of every frame, the BSbroadcasts a notication message to all number of devices tonotify the beginning of the frame.Upon receiving the noticationmessage, the active devices prepare to contend the transmissiontime slots. Other devices that do not have packets to send willenter sleep mode to preserve energy. By knowing packet arrivalrate of every device, the BS estimates the number of activedevices and calculates the optimal contending parameters: con-tending duration , initial contending probability ,and incremental indicator by solving an optimization prob-lem. The specic description and solution of the optimizationproblem is described in Section IV. These contending parametersare included in the notication message and broadcasted by BSduring NP. Upon receiving the notication message, the activedevices will calculate their own contending probabilities. Thecalculation is based on an incremental contending prioritymechanism which will be presented in Section III-B. Next, theM2M network enters to COP.2) Contention Only Period: In this period, we, respectively,

present the operations of devices and BS as follows.Devices: In this period, devices contend the transmission

opportunities based on -persistent CSMA mechanism, accord-ing to their own contending probability . The contendingdevices randomly send the transmission request (Tran-REQ)message to the BS. The contention is declared as success only

when one device sends the Tran-REQmessage. Whenmore thanone devices are sending Tran-REQduring the same time interval,a collision occurs. The idle period is a time interval in which thecontention is not happening. Under -persistent CSMA, thesuccess period and collision period can be given as

and

where is the length of Tran-REQ message, is theduration of acknowledge (ACK) message, and andare the backoff inter-frame space and short inter-frame space,respectively.Upon receiving the ACK message from BS, the device will

stop sending Tran-REQ message and waits for the AP duration.In addition, the ACK message includes the information of theindex of the transmission time slot that the device is allowed totransmit in TOP.BS: If a Tran-REQ message is successfully received from a

device, the BS sends ACK message and the index of thetransmission time slot to this device. For a given optimalcontention period , we may have more number ofdevices successfully contend a time slot than that is allowablefor a given frame size (the expected number of successful devicesduring is denoted as ). Hence, andare used as two thresholds to control the duration of COP inpractice. When the actual number of successful devices incontention is greater than or the is longer than

, the BS will stop the COP and declare the next period,i.e., AP.3) Announcement Period:At the beginning of AP, BS initiates

and broadcasts the announcement message to all of the devices.Upon receiving the announcement message, the devicessucceeded in contention turn to transmission mode andprepare to send their own packets. Other active devices thatfailed in contention cease their contending operations and turn tosleepmode. Such arrangement keeps thewakeup time of a deviceat minimal, and we will further study the energy consumption ofthe proposed protocol in Section V.4) Transmission Only Period: In TOP, active devices

succeeded in contention sequentially operate the transmissionfollowing the TDMA mechanism. These devices turn ON theirradio modules and transmit the data packet during their owntransmission time slots. The devices turn OFF the radio module atother time slots. When the timer of TOP is out, the BS declaresthe beginning of a new frame. Although only uplink ismentioned, some modication to the protocol can be applied

Fig. 2. Frame structure.

LIU et al.: DESIGN OF A SCALABLE HYBRID MAC PROTOCOL 101

to downlink where the devices would like to receive informationfrom the BS.It is expected that higher number of devices succeeded in

contention can be obtained if gets longer. However, giventhe duration of a frame , the will be decreased as the

increases, whichmay reduce the total transmission time forallowing successful devices to transmit data. Hence, there is atradeoff between the durations of COP and TOP. To balance thistradeoff, we intend to propose a hybrid access control scheme,presented in Section IV, which focuses on obtaining the optimal

, , and to maximize the channel utility.

B. Incremental Contending Priority Mechanism for Fairness

According to -persistent CSMAmechanism, the devicesmayfail in the contention during COP and lose the transmissionopportunities at TOP. If a device frequently fails in contentionframe-by-frame, the transmission performance of this devicewilldrastically degrade. Hence, we propose an incremental conten-tion priority model for the sake of fairness. In this model, weincrease the contending probability of the device frame-by-frameif such device failed to contend the transmission opportunities atprevious frames. When the device successfully transmit a datapacket, the increasing process will be stopped and the priority ofthe device will return to the preliminary level. For type devicesthat have a new packet arrival, the BS assigns the preliminarycontending probability . If the type devices failed in conten-tion at previous frames, the BS will increase their contendingprobability at current frame according to

where is the number of frames during which thedevices failed in contention and is the incremental indicator.For simplicity, in this paper, we assume that the contendingprobability dened in (1) and (2) has the same incrementalindicator (in general, they can be two different parameters).

Hence, we note that there could be more than one class ofdevices that have the same contending probabilities at a certainframe. For example, there will be three kinds of devices that havethe same contending probability at a frame: the class 3 deviceshave preliminary contending probability ; theclass 2 devices, which failed once in previous contention, haveincreased the contending probability to

; and the class 1 devices, which failed continuouslyacross two frames, have increased the contending probability to

. For easy expression, we dene the virtualclass in which the devices, at a certain frame,have the same contending probability . Then, we have

where and.

C. An Illustrative Example

An example of using the proposed protocol is illustrated inFig. 3, where the hybrid access operations of eight devices areconsidered. To clearly describe our protocol, we only considerone type of priority class in the example, i.e., , then,

. The operation of each device in theM2M network includes two processes: 1) contention and trans-mission process and 2) new packet arrival process. At frame 0,there is no contention and transmission process for any devices.We can see that the active devices are , , , and at theend of frame 0. Note that during frame 0, has two packetsarrival and the former arrived packet will be replaced by the latterone. At frame 1, , , , and take part in contention withthe contending probability . and , which succeeded incontention are allowed to transmit data during the followingTOP. and , which failed in contention, should wait forcontention again in the next frame. Meantime, their contendingprobabilities increase from to , where . Inaddition, we can see that , , , , and have new

Fig. 3. Frame-based hybrid access process.


packet arrival in frame 1. Hence, the devices will contend inframe 2 are , , , , , and , where the contendingprobability of and is and that of the rest is . Afterframe 2, we can see that failed in contention again and itscontending probability increases to , where .Finally, succeed in frame 3 and also has a new packet arrivalduring frame 3. will go back to the contention at frame 4 withcontending probability .

IV. HYBRID MULTIPLE ACCESS CONTROL INMASSIVE M2M NETWORK

In this section, we rst provide the expression of averageduration of COP T in terms of the number of successfuldevices , initial contending probability , and incrementalindicator . Then, we formulate an optimization problem tomaximize the channel utility. The ofine solution of the optimi-zation problemwill be given to nd the optimal , ,and .

A. Derivation of Average

According to our proposed MAC protocol, the active devicesin virtual class have the contending probability and willcontend the transmission slots during COP in a frame. To obtainthe expression of , we need to know the average contendingtime of each device. First, we make the following assumptionsand notations:1) The packet arrival rate ( ) at each frame and the duration of

each frame ( ) are constant.2) For frame , there are virtual classes of

devices, each of which contains devices ( ),

where , is the

number of th class of devices that succeeded in

contention at th frame, is the number ofempty class devices that have new packet arrival during

th frame.3) A class device uses the probability in

the -persistent CSMA ( > ) to contendthe transmission opportunities.

Based on -persistent CSMA in COP period, when a conten-tion attempt is completed (successfully or with a collision), theactive device will start a contention attempt with probability

. Here, we dene the successful contention asthe event that the transmission request from a device is success-fully received by BS. Let denote the time between the( )th and the th successful contentions at frame . Let

denote the number of collisions that occur during , then

where is the duration of the th idle time that precedesthe channel busy period (either collision or success) in each

duration. is the duration of the th collision giventhat a collision occurs, and is the length of the request

message. Let denote the average contention time at frame, then, we have

where , , , and are the averagenumber of collisions, the average duration of an idle time, acollision, and a request message at frame , respectively.The total number of devices that can successfully contend the

transmission opportunities at frame is denoted by . Letdenote the duration of the COP at steady frame . Then,we

have

Since is the sum of random variable, the is also a random variable with the

average time for number of successful contentions. Toobtain the close-form expression of the , we then focus onderiving the expected value of , which is denoted by

. Due to the independently distributed , we have

Then, we derive , , and in the case ofmultiple priority classes.Derivation of : Let denote the number of contend-

ing devices in a contending slot immediately after an idleinterval, and let and , respectively, denote theprobability that a collision occurs and that a transmission issuccessful, both conditioned on that at least one device transmitthe contending message. Then,

and

The probability distribution of can be expressed as


and can be obtained as

Derivation of : Since class devices may contend in aslot with probability , we have

where is a constant [29].Let and . Then, is the

function of and . Let T , aftersome algebraic manipulations:

T

In Section VI-B, we use T to expressT , since , where

.

B. Optimization Problem Formulation

Given , longer T allows more devicesto succeed in contention. However, the incrementalT will reduce the duration of TOP subjectingto the constraint as T . Tobalance this tradeoff, we formulate an optimization problem tomaximize the channel utility in each frame. In this paper, thechannel utility, denoted by C, is dened as the mean value of theduration of TOP divided to the entire duration of frameover frames

C

where denote the duration of a transmission time slot. Then,we can maximize the channel utility as follows:

T

C

T

< >

The derivation of in constraint (17) isgiven in Appendix A. In constraint (18), is dened as theprobability that a class device successfully contend in thecontending time

Hence, the number of class devices successfully contend thetransmission opportunities at frame is given by

Note that T is the function of as shown in (14). Forsimplicity of expression, we use variable instead of T .Finally, the previous optimization problem can be written as

Next, we try to prove the convexity of the above-mentionedoptimization problem. It is easy to observe that the objectivefunction (15) is a linear function of and constraints (17)(19)are linear. For constraint (16), we provide the convexity proof byTheorem 1.

Theorem 1: Let , for and ,T can be obtained as a convex function of

, , and .

Proof: The proof is presented in Appendix B. Theorem 1 shows that, asymptotically, forM2Mnetworkswith

tremendous number of devices, i.e., when the number of activedevices is large, constraint (16) is also a convex function. There-fore, the optimization problem is a convex programming problemand can be solved easilywith off-the-shelf toolbox and the optimal

period of COP, T .


V. PERFORMANCE STUDY AND EVALUATION

In this section, we analyze channel utility, packet drop ratio,average transmission delay, and energy consumption of theproposed hybridMACprotocol in heterogeneousM2Mnetwork,which consist of ( ) number of devices.Without loss of generality, in our simulation, the heterogeneousM2M network includes three classes of devices: class 1 (device

), class 2 (device ), and class 3 (device), which have the contending probabilities

, , and , respectively.Meanwhile, the performance comparisons of the proposed pro-tocol with contention-based protocol -persistent CSMA [29]and reservation-based protocolTDMA [22] are provided.Summarily, the simulation parameters are shown in Table I.

A. Channel Utility

Fig. 4 shows the channel utility with different packet arrivalrates, in terms of the initial contending probability in 500(straight line), 800 (circle), and 1200 (diamond) devices cases. Itis observed that the channel utility increases at rst and thendecreases as the increases. This is because the number ofsuccessful devices in contention increases as the contending

probability increases for xed total number of devices. Accord-ingly, the transmission of the successful devices will lead toincreasing channel utility of a frame. However, as the contendingprobability increases to or exceeds a certain value, the number ofcontending collisions may become large and the duration of thecontention may increase, which results in the decrement oftransmission period (TOP).Table II shows the channel utility in 500, 800, and 1200

devices cases when , in terms of the initial contendingprobability , and incremental indicator . From Table II, weobserve that the channel utility ismaximized as 0.6229 (and ), 0.6760 ( and ), and0.7888 ( and ) in 500, 800, and 1200devices cases, respectively. The channel utility in the smallnumber of devices case is lower than that in the large numberof devices case. This is expected since the channel utilizationmay be reduced in the small number of devices condition due tolimited transmission requirements through the entire network.Fig. 5 shows the channel utility comparison in terms of packet

arrival rate among the proposed hybrid protocol, CSMA, andTDMA protocols in 1200 devices case. In this case, the proposedhybrid protocol chooses the optimal parameters for devices as

and . It is observed that the channel utilityin the proposed hybrid protocol isrst higher and then lower thanTDMA, and will be higher than -persistent CSMA as packetarrival rate increases. That is, the proposed hybrid protocolcan optimally control the initial contending probability ,incremental indicator , and the number of successful devices tomaximize the channel utility. While -persistent CSMA mayperform well only at low-load condition, TDMA performs wellonly at heavy-load condition. Our hybrid protocol outperforms

TABLE ISIMULATION PARAMETERS

Fig. 4. Channel utility with different packet arrival rates in terms of whenin 500 (straight line), 800 (circle), and 1200 (diamond) cases.

TABLE IIUTILITY IN TERMS OF AND WHEN


the TDMAwhen is small and performs better than -persistentCSMA when is increased.

B. Packet Drop Ratio

In this paper, we use packet drop ratio as a metric to measurethe network throughout. The packet drop ratio is dened as theratio of the number of dropped packets to the total number ofgenerated packets per device during the simulation duration.Fig. 6 shows the packet drop ratio of the proposed hybridprotocol in 1200 devices case when and .We observe that the packet drop ratio of the proposed protocolis rst higher and then lower than TDMA and -persistentCSMA as packet arrival rate increases. That is, the proposedhybrid protocol is able to increase the contending probabilityfor the device when it continually failed in the contention inprevious frames. This mechanism can guarantee that all devices

have fair chance to obtain the transmission slot, which leadsto low packet drop ratio. Without this mechanism, the TDMAand -persistent CSMA may perform well only at low-loadcondition. When packet arrival rate is increasing, the packetdrop ratio of both of TDMA and -persistent CSMA will bedrastically increasing.In Section V-B-1, we focus on analyzing the packet drop ratio

of the proposed hybrid protocol in both of the homogeneous andheterogeneous M2M networks, which consist of 1200 devices.1) Homogeneous Case: Different from the setting in

heterogeneous case, all devices have the same contendingprobability in homogeneous case. Fig. 7 shows thepacket drop ratio of the homogeneous network with

and in 1200 homogeneous devicescase when and . We observe that thepacket drop ratio of the devices is nearly the same in 1 pac/scase (between 0.3 and 0.4) and 2 pac/s case (between 0.55 and0.65), respectively. That is, based on our protocol, the devicesthat failed in contention at current frame will be given highercontending priorities in the following frame. This adaptiveadjustment can guarantee that all of the devices in M2Mnetwork have the fair opportunity to compete the transmissiontime slots.2) Heterogeneous Case: Fig. 8 shows the packet drop ratio

when packet arrival rate is and inheterogeneous 1200 devices case when and

. We observe that class 3 devices have the lowestpacket drop ratio and class 1 devices have the highest packetdrop ratio on an average. This indicates that the differentcontending probabilities have big inuence on theperformance of the different classes devices in heterogeneousM2M networks.

C. Average Transmission Delay

In this section, we dene the average transmission delay as theaverage number of frames during which the active devicesuccessfully nishes a transmission. Suppose that a device is

Fig. 5. Channel utility comparison in terms of packet arrival rate in 1200devices case when and .

Fig. 6. Comparison of packet drop ratio in terms of packet arrival rate ( ) in 1200devices case when and .

Fig. 7. Packet drop ratio with packet arrival rates (1 pac/s and 2 pac/s) in 1200homogeneous devices case when and .


activated (new packet arrival) at frame and successfullynishes its transmission at frame , where

. Then, the average transmission delay of a device canbe obtained by

where is the number of successful transmissions of the deviceduring simulation. In simulation, we run the operation of theM2M network over frames.The average transmission delay in 500, 800, and 1200 devices

caseswith packet arrival rate (1 pac/s) when andare shown in Fig. 9.We can observe that the average transmissiondelay increases as the number of devices increases. This isbecause the growing number of devices will cause the increasingcollisions at COP of a frame, which leads to the increasingtransmission delay.Fig. 10 shows the comparison of average transmission

delay among proposed hybrid method, -persistent CSMA,

and TDMA in 500, 800, and 1200 devices cases. In each case,the packet arrival . For the proposed hybrid protocol,

and . The comparison indicates that the pro-posed hybrid protocol is able to obtain less transmission delaythan -persistent CSMA and TDMA in heavy load case (800and 1200 devices cases). This is because the hybrid protocol canoptimally control the contention period to allowmore devices tohave the transmission opportunities when the number of de-vices becomes very large.Next, we also show the average transmission delay of the

proposed hybrid protocol in both of the homogeneous andheterogeneous M2M networks, where the number of devices is1200. Similar to the packet drop ratio evaluation, we rst select60 derives among the 1200 devices to show the simulationresults.1) Homogeneous Case: In homogeneous case, all devices

have the same contending probability . Fig. 11 shows theaverage transmission delay of the devices 160 when the packetarrival rate of each devices are 1 and 2 pac/s, respectively. It is

Fig. 8. Packet drop ratio with packet arrival rates (1 pac/s and 2 pac/s) inheterogeneous 1200 devices case when and .

Fig. 9. Average transmission delay with packet arrival rate (1 pac/s) in 500, 800,and 1200 devices case when and .

Fig. 10. Comparison of average transmission delay in terms of the total numberof devices.

Fig. 11. Average transmission delaywith packet arrival rates (1 pac/s and 2 pac/s)in 1200 devices case when and .


observed that the delay of each device uctuate stably centeredon 0.4 frames in 1 pac/s case and 1.6 frames in 2 pac/s case. Thisindicates that our proposed hybrid protocol can guarantee thefairness of the devices to obtain the transmission opportunities.2) Heterogeneous Case: Fig. 12 shows the average

transmission delay of all 60 devices when the packet arrivalrates of each device are 1 and 2 pac/s, respectively. As expected,it is shown that the class 3 devices have the lowest transmissiondelay and class 1 devices have the highest average delay.

D. Energy Consumption

In this section, we consider the energy consumption of theM2M network in one frame. In this paper, we consider that BSobtains power from grid and the devices obtain power frombattery. Hence, we focus on the power consumption of thedevices. For a device, the power consumption during the trans-mission mode is denoted by P W; the power consumptionduring the receiving mode is denoted by P W; and the powerconsumption during the idle mode is denoted by P W. Theenergy consumption of the M2M network in each period isdened as follows.During NP duration, each device receives a notication

message. Let denote the total energy used for receivingthe notication messages during NP. Then, we have

P

where is the number of all devices inM2M network andis the length of the notication message.During COP duration, there will be devices that suc-

cessfully send Tran-REQ message to BS. The total energy usedfor sending this message is denoted by , we have

where P P PP .

During AP duration, the BS broadcasts an announcementmessage to all active devices to announce the end of COP andthe start of TOP. Hence, number of devices receive themessage and let denote the total energy used for receivingthe announcement message from the BS during AP

P

where is the length of the announcement message.During TOP duration, after receiving the allocated transmis-

sion schedule from the BS, each device sends its data packet tothe BS at its scheduled time slots . The energy consumption bysuccessful devices in transmission during a single frame isdened as follows:

P

The devices that failed in contention stay idle and keep theirradio module OFF during TOP. Thus, over a single frame, itconsumes the following energy:

P

where P denotes the energy used in idle mode duringTOP. Hence, the total energy consumed by the devices duringTOP is

Therefore, the total energy consumption of all devices duringone frame is dened as follows:

In simulation, we aim to compare the energy consumption in aframe among the proposed hybrid protocol, -persistent CSMAand TDMA. The packet arrival rate of each device is . Forthe proposed hybrid protocol, and . Fig. 13

Fig. 12. Average transmission delaywith packet arrival rates (1 pac/s and 2pac/s)in heterogeneous 1200 devices case when and . Fig. 13. Comparison of energy consumption in terms of the total number of

devices.


shows the comparison of energy consumption among theproposed hybrid method, -persistent CSMA and TDMA. Thecomparison indicates that the proposed hybrid protocol is ableto consume less energy than -persistent CSMA. This isbecause the hybrid protocol only allows the devices to transmita small length contending message during contention period.The energy consumption during collision can be greatlyreduced. Moreover, when the contention is nished, the pro-posed hybrid method can control the devices that failed incontention turn to the idle mode for saving energy. In addition,the proposed protocol consumed more energy compared to thatof TDMA scheme. That is, in the proposed protocol, deviceshave to use more energy for contending during the contentionperiod; however, this energy consumption can lead to higherchannel utility and lower packet drop ratio as shown in Sec-tions V-A and V-B.

VI. CONCLUSION

In this paper, we focused on designing the massive MACprotocol for heterogeneous M2M network where the deviceshave different service requirements. In our protocol, the opera-tion of each frame is mainly divided into two parts: 1) COP and2) TOP. The heterogeneous devices with different contendingprobability contend the transmission time slots during COP andonly the successful devices in contention will be assigned thetime slots for transmission. Considering the fairness, the con-tending probability of the device that failed in contention atprevious frame will be increased at the next frame. Under suchmechanism, the BS can easily maximize the channel utility bycontrolling the duration of COP , initial contending proba-bility , and the incremental indicator . An optimizationproblem was formulated to solve the problem, and we showedanalytically that the problem is convex.We analyzed the channelutility, packet drop ratio, average transmission delay, and energyconsumption to show the effectiveness of the propose hybridMAC protocol; especially, there are heterogeneous devices withdifferent priorities.

APPENDIX A

DERIVATION OF

is dened as the number of empty classdevices that have new packet arrival during th frame.Recall that the packet arrival process is a Possion arrival processwith arrival rate at each device. Let denote the probabilitythat a type device has at least one new packet arrival during

. Then, we have

Next, we rst calculate when . In this case,denote the number of empty class devices that have newpacket arrival during frame 0. In frame 0, the vitual priorities ofdevices are only decided by the heterogeneous type in M2Mnetworks; hence, . Let representthe number of class devices that have at least one new packet

arrival during frame 0. We can obtain the probability thatas

Therefore, we can obtain

For , the number of empty type devices

, where and

is given by (20). Let denote the number of emptyclass devices that have at least one new packet arrival during

th frame and denote the probability that

, we have

Similarly, we can calculate

APPENDIX B

PROOF OF THEOREM 1

Since the duration of has a nite value, as , it iseasy to obtain , then we have

T

Moreover, tends to ifis sufciently large. Hence, we can obtain the approximated

transformation of the above-mentioned equation as

T


Taking the second derivative of T withrespect to , , and , respectively, the Hessian matrix isgiven by

T T T

T T T

T T T

Recall that , it is easy to obtain

T

T

>

T

>

T T

>

T T

>

T T

>

Consequently, the Hessian matrix of T , , we

conclude that T is a convex function of, , and [30].

REFERENCES

[1] M. Stanley, Internet trends, (Mar. 9, 2010) [Online]. Available: http://bbh-labs.com/internet-trends-2010-by-morgan-stanleys-mary-meeker

[2] IEEE C80216-10_0002r7, Machine to machine (M2M) communicationstudy report, IEEE 802.16 Contribution, May 2010.

[3] G. Wu, S. Talwar, K. Johnsson, N. Himayat, and K. D. Johnson, M2M:From mobile to embedded internet, IEEE Commun. Mag., vol. 49, no. 4,pp. 3643, Apr. 2011.

[4] J. Chen, X. Cao, P. Cheng, Y. Xiao, and Y. Sun, Distributed collaborativecontrol for industrial automationwithwireless sensor and actuator networks,IEEE Trans. Ind. Electron., vol. 57, no. 12, pp. 42194230, Dec. 2010.

[5] Y. Zhang, R. Yu, S. Xie, W. Yao, Y. Xiao, and M. Guizani, Home M2Mnetworks:Architecture, standards andQoS improvements, IEEECommun.Mag., vol. 49, no. 4, pp. 4452, Apr. 2011.

[6] R. Yu, Y. Zhang, S. Gjessing, W. Xia, and K. Yang, Toward cloud-basedvehicular networkswith efcient resourcemanagement, IEEENetw.Mag.,vol. 27, no. 5, pp. 4855, Sep./Oct. 2013.

[7] X. Cao, J. Chen, Y. Zhang, and Y. Sun, Development of an integratedwireless sensor networkmicro-environmentalmonitoring system,ElsevierISA Trans., vol. 47, no. 3, pp. 247255, 2008.

[8] R.-H. Gau and C.-P. Cheng, Optimal tree pruning for location update inmachine-to-machine communications, IEEE Trans. Wireless Commun.,vol. 12, no. 6, pp. 26202632, Jun. 2013.

[9] S.-Y. Lien, T.-H. Liau, C.-Y. Kao, and K.-C. Chen, Cooperative accessclass barring for machine-to-machine communications, IEEE Trans.Wireless Commun., vol. 11, no. 1, pp. 2732, Jan. 2012.

[10] C.-Y. Tu, C.-Y. Ho, and C.-Y. Huang, Energy-efcient algorithms andevaluations for massive access management in cellular based machine tomachine communications, in Proc. IEEE Veh. Technol. Conf. (VTCFall), 2011, pp. 15.

[11] Y. Liu, S.-L. Xie, R. Yu, Y. Zhang, and C. Yuen, An efcient MACprotocol with selective grouping and cooperative sensing in cognitive radionetworks, IEEE Trans. Veh. Technol., vol. 62, no. 8, pp. 39283941,Oct. 2013.

[12] R.Yu, C. Zhang,X. Zhang, L. Zhou, andK.Yang, Hybrid spectrum accessin cognitive radio based smart grid communications networks, IEEE Syst.J., to be published.

[13] Machine-to-machine (M2M) & smart systems forecast 20102014,Harbor Research Rep., (2009) [Online]. Available: http://www.reportlinker.com/p0210447/Machine-To-Machine-M2M-Smart-Systems-Forecast.html

[14] 3GPP TS 22.368 V11.1.0, (2011). Service Requirements for Machine-TypeCommunications [Online]. Available: http://www.3gpp.org/ftp/Specs/archive/22_series/22.368/22368-b10.zip

[15] 3GPP TR 23.888 V1.2.0, (2010). System Improvement for Machine-TypeCommunications [Online]. Available: http://www.3gpp.org/ftp/Specs/archive/23_series/23.888/23888-120.zip

[16] IEEE Standard for Local and Metropolitan Area NetwroksPart 16: AirInterface for BroadbandWireless Access Systems Amendment 3: AdvancedAir Interface, IEEE 802.16 m-2011, 2011.

[17] Enhancements to Support Machine-to-Machine Applications, IEEE802.16p AWO, 2011.

[18] Y. Zhang, R. Yu, M. Nekovee, Y. Liu, S. Xie, and S. Gjessing, Cognitivemachine-to-machine communications: Visions and potentials for the smartgrid, IEEE Netw., vol. 26, no. 3, pp. 613, May/Jun. 2012.

[19] M. Hasan, E. Hossain, and D. Niyato, Random access for machine-to-machine communication in LTE-advanced networks: Issues andapproaches, IEEE Commun. Mag., vol. 51, no. 6, pp. 8693, Apr. 2013.

[20] G. Wang, X. Zhong, S. Mei, and J. Wang, An adaptive medium accesscontrol mechanism for cellular based machine to machine (M2M) com-munication, in Proc. IEEE Int. Conf. Wireless Inf. Technol. Syst.(ICWITS), 2010.

[21] S. Kim, J. Cha, S. Jung, C. Yoon, and K. Lim, Performance evaluation ofrandomaccess forM2Mcommunication on IEEE802.16network, inProc.ICACT, 2012, pp. 278283.

[22] T.-H. Hsu and P.-Y. Yen, Adaptive time division multiple access-basedmedium access control protocol for energy conserving and data transmis-sion in wireless sensor networks, IET Commun., vol. 5, no. 18,pp. 26622672, 2011.

[23] I. Rhee, A. Warrier, M. Aia, J. Min, and M. L. Sichitiu, Z-MAC: A hybridMAC for wireless sensor networks, IEEE Trans. Netw., vol. 16, no. 3,pp. 511524, Jun. 2008.

[24] R. Zhang, R. Ruby, J. Pan, L. Cai, and X. Shen, A hybrid reservation/contention-based MAC for video streaming over wireless networks, IEEEJ. Sel. Areas Commun., vol. 28, no. 3, pp. 389398, Apr. 2010.

[25] G. Bianchi, Performance analysis of the IEEE 802.11 distributed coordi-nation function, IEEE J. Sel. Areas Commun., vol. 18, no. 3, pp. 535548,Mar. 2000.

[26] M. Heusse, F. Rousseau, R. Guillier, and A. Duda, Idle sense: An optimalaccess method for high throughput and fairness in rate diverse wirelessLANs, in Proc. SIGCOMM05, Aug. 2126, 2005.

[27] D. Malone, K. Duffy, and D. Leith, Modeling the 802.11 distributedcoordination function in non-saturated heterogeneous conditions, IEEE/ACM Trans. Netw., vol. 15, no. 1, pp. 159172, Feb. 2007.

[28] R. K. Lam and K.-C. Chen, Congestion control for M2M trafc withheterogeneous throughput demands, in Proc. IEEE Wireless Commun.Network. Conf. (WCNC), 2013.

[29] R. Bruno, M. Conti, and E. Gregori, Optimal capacity of p-persistentCSMAprotocols, IEEECommun.Lett., vol. 7, no. 3, pp. 139141,Mar. 2003.

[30] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge, U.K.:Cambridge Univ. Press, 2004.

[31] L. Kleinrock, Queuing System. New York, NY, USA: Wiley, 1975.


Yi Liu received the Ph.D. degree from South ChinaUniversity of Technology (SCUT), Guangzhou,China, in 2011.After that, he worked with the Institute of Intelli-

gent Information Processing, Guangdong Universityof Technology (GDUT), Guangdong, China. In 2011,he joined as a Postdoctoral with the SingaporeUniversity of Technology andDesign, Singapore. Hisresearch interests include cognitive radio networks,cooperative communications, smart grid and intelli-gent signal processing.

Chau Yuen (S02M08SM12) received theB.Eng. and Ph.D. degrees from Nanyang Tech-nological University, Singapore, in 2000 and 2004,respectively.He was a PostDoc Fellow at Lucent Technologies

Bell Labs (Murray Hill) during 2005, and a VisitingAssistant Professor with Hong Kong PolytechnicUniversity, Hong Kong, in 2008. From 2006 to2010, he worked as a Senior Research Engineer withthe Institute for Infocomm Research, Singapore. Hejoined as an Assistant Professor with Singapore

University of Technology and Design in June 2010. He has published over150 research papers in international journals or conferences. His current researchinterests include green communications, massive MIMO, Internet-of-things,machine-to-machine, network coding, and distributed storage.Dr. Yuen also serves as an Associate Editor for the IEEE TRANSACTIONS ON

VEHICULAR TECHNOLOGY.

Xianghui Cao (S08M11) received the B.S. andPh.D. degrees in control science and engineeringfrom Zhejiang University, Hangzhou, China, in2006 and 2011, respectively.From 2007 to 2009, he was a visiting scholar with

the Department of Computer Science, University ofAlabama, Tuscaloosa,USA.He is anAssociate Editorfor KSII Transactions on Internet and InformationSystems and Security and Communication Networks(Wiley). His research interests include wirelessnetwork performance analysis, energy efciency of

wireless networks, networked estimation and control, and network security.Dr. Cao is a TPC member for IEEE Globecom 2013, 2014, IEEE ICC 2014,

IEEE VTC 2013, 2014, etc.

Naveed Ul Hassan received the B.E. degree in avi-onics engineering from the College of AeronauticalEngineering, Risalpur, Pakistan, in 2002, theMastersand Ph.D. degrees in telecommunications from EcoleSuperieure dElectricite (Supelec), Gif-sur-Yvette,France, 2006 and 2010, respectively.Since August 2011, he serves as an Assistant

Professor with the Department of Electrical Engineer-ing, Lahore University of Management Sciences(LUMS), Lahore, Pakistan. His research interestsinclude cross layer design and optimization in wire-

less networks, heterogeneous networks, cognitive radio networks and smart grids.

Jiming Chen (M08SM11) received the B.Sc. andPh.D. degrees both in control science and engineeringfrom Zhejiang University, China, in 2000 and 2005,respectively.He was a Visiting Researcher with Institut National

de Recherche en Informatique et en Automatique(INRIA), France, in 2006, National University ofSingapore, Singapore, in 2007, and University ofWaterloo, Waterloo, QC, Canada, from 2008 to2010. Currently, he is a Full Professor with theDepartment of Control Science and Engineering, the

Coordinator of Group of Networked Sensing and Control in the State KeyLaboratory of Industrial Control Technology, and Vice Director of the Instituteof Industrial Process Control, Zhejiang University, Zhejiang, China.Dr. Chen currently serves as an Associate Editor for several international

journals including IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEM,IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, IEEE NETWORK, IEEE TRANS-ACTION ON CONTROL OF NETWORK SYSTEMS, etc. He was a Guest Editor of IEEETRANSACTIONS ON AUTOMATIC CONTROL, Computer Communication (Elsevier),Wireless Communication andMobile Computer (Wiley) and Journal of Networkand Computer Applications (Elsevier). He also served/serves as Ad hoc andSensor Network Symposium Co-chair, IEEE Globecom 2011; general symposiaCo-Chair of ACM IWCMC 2009 and ACM IWCMC 2010, WiCON 2010MAC track Co-Chair, IEEE MASS 2011 Publicity Co-Chair, IEEE DCOSS2011 Publicity Co-Chair, IEEE ICDCS 2012 Publicity Co-Chair, IEEE ICCC2012 Communications QoS and Reliability Symposium Co-Chair, IEEESmartGridComm The Whole Picture Symposium Co-Chair, IEEE MASS2013 Local Chair, Wireless Networking and Applications Symposium Co-chair,IEEE ICCC 2013, Ad hoc and Sensor Network Symposium Co-chair, IEEEICC 2014, and TPC member for IEEE ICDCS10,12,13,14, IEEEMASS10,11,13, IEEE SECON11,12, IEEE INFOCOM11,12,13,14, etc.


1405.6360.pdf

Documents

machine m2m network

m2m communications

m2m networkthat

different devices

devices andsystems

successful devices

geneous devices

arange of devices